Comparing higher order models for the EORTC QLQ-C30

Abstract

Purpose: To investigate the statistical fit of alternative higher order models for summarizing the health-related quality of life profile generated by the EORTC QLQ-C30 questionnaire.

Methods: A 50% random sample was drawn from a dataset of more than 9,000 pre-treatment QLQ-C30 v 3.0 questionnaires completed by cancer patients from 48 countries, differing in primary tumor site and disease stage. Building on a "standard" 14-dimensional QLQ-C30 model, confirmatory factor analysis was used to compare 6 higher order models, including a 1-dimensional (1D) model, a 2D "symptom burden and function" model, two 2D "mental/physical" models, and two models with a "formative" (or "causal") formulation of "symptom burden," and "function."

Results: All of the models considered had at least an "adequate" fit to the data: the less restricted the model, the better the fit. The RMSEA fit indices for the various models ranged from 0.042 to 0.061, CFI's 0.90-0.96, and TLI's from 0.96 to 0.98. All chi-square tests were significant. One of the Physical/Mental models had fit indices superior to the other models considered.

Conclusions: The Physical/Mental health model had the best fit of the higher order models considered, and enjoys empirical and theoretical support in comparable instruments and applications.

Fig. 1

Seven hypothesized models^a: a standard model, b physical health, mental health and QL, c physical burden, mental function and QL, d symptom burden, function and QL, e HRQL and QL, f formative symptom burden (free weights), function and QL, g formative symptom burden (fixed weights) function and QL. ^aModels are described in text. Item thresholds, means, (error) variances, and correlations between first-order latent variables (in the standard model) are not represented, for clarity’s sake